Cosmin Safta

Cosmin Safta

Distinguished Member of Technical Staff

Sandia National Laboratories

Bio

I am a Distinguished Member of Technical Staff at Sandia National Laboratories. My research focuses on computational science with 20+ years of experience in developing algorithms for a large set of applications. I am involved in and leading research projects pertaining to uncertainty quantification and machine learning in combustion, climate modeling, biosurveillance, power grids, material science, stochastic network models, and re-entry vehicle trajectories, as well as statistical modeling related to adversarial aspects in machine learning.

Download my resumé.

Interests
  • Uncertainty Quantification
  • Machine Learning
  • Bayesian Inference
Education

  • PhD in Mechanical and Aerospace Engineering, 2004

    University of Buffalo

  • MSc in Aerospace Engineering, 1996

    University "Politehnica" of Bucharest

  • BSc in Aerospace Engineering, 1995

    University "Politehnica" of Bucharest

Experience

 
 
 
 
 
Senior/Principal/Distinguished Member of Technical Staff
Sandia National Laboratories
Jul 2009 – Present California
  • Developing state-of-the-art algorithms for uncertainty quantification and scientific machine learning for large scale computational models of physical phenomena
  • Deploying software for various tasks in model development and data analysis pipeline, including uncertainty propagation, model calibration and statistical analysis
  • Leading and participating in research proposals and scientific projects
  • Hiring and mentorship of postdoctoral appointees and summer students
 
 
 
 
 
Postdoctoral Researcher
Sandia National Laboratories
Jan 2007 – Jul 2009 California
High-order adaptive mesh refinement algorithms for chemical reacting flows. Designed and published the first high-order discretization scheme for low-speed combustion applications.
 
 
 
 
 
Postdoctoral Researcher & Research Engineer
SUNY Stony Brook and TTC Technologies
Oct 2004 – Dec 2006 Long Island, NY
Algorithms for turbulence-combustion coupling in reacting flow simulations. Algorithm developments were transitioned to the Air Force Research Laboratory.

News

Co-organizing minisymposium on "Transfer Learning and Multi-Fidelity Approaches to Alleviate Data Sparsity in Machine Learning" at SIAM-MDS22
https://meetings.siam.org/program.cfm?CONFCODE=mds22
Cosmin Safta
Jun 1, 2022 minisimposium
Co-organizing minisymposium on "Incorporating fundamental principles in innovative machine learning models of physics" at WCCM22
https://www.wccm2022.org
Cosmin Safta
Mar 14, 2022 minisimposium
Co-organizing minisymposium on "Uncertainty Quantification and Propagation in Climate Models" at SIAM-UQ22
https://www.siam.org/conferences/cm/conference/uq22
Cosmin Safta
Mar 1, 2022 minisimposium
Submitted white paper to AI4ESP
White papers were solicited for development and application of AI methods in areas relevant to EESSD research with an emphasis on quantifying and improving Earth system predictability, particularly related to the integrative water cycle and extreme events. Our white paper focuses on improving the earth system model through uncertainty attribution and active learning.
Cosmin Safta
Apr 1, 2021 white paper
Sandia National Labs' 2020 HPC Annual Report Released
Here is a short video describing our project (link).
Cosmin Safta
Mar 1, 2021 white paper

Projects

Graph Neural Network Models
Graph Neural Network Models of Complex Initial Boundary Value Problems that embed Physical Invariances
Graph Neural Network Models
Tensor and Bayesian Framework for Machine Learning
Functional tensor network models.
Tensor and Bayesian Framework for Machine Learning

Recent & Upcoming Talks

Machine learning constitutive models of inelastic materials with microstructure
Graph-based neural networks designed to model the response of elastic-plastic and viscoelastic materials with pores.
Jul 31, 2022 12:00 AM — 12:00 AM Yokohama, Japan
Development of a spatially regularized detector for emergent/re-emergent disease outbreaks
Bayesian framework to detect outbreaks of emerging diseases.
Jun 26, 2022 12:00 AM — 12:00 AM Montreal, Canada
Quantifying Uncertainty in E3SM via Functional Tensor Network Approximations
Employ low-rank functional tensor network models for UQ in earth system models.
Apr 12, 2022 12:00 AM — 12:00 AM Atlanta, GA
Functional Tensor Network Approximations for E3SM Land Model
Employ low-rank functional tensor network models for global sensitivity analysis at a regional scale.
Dec 13, 2021 12:00 AM — 12:00 AM New Orleans, LA
Machine learning constitutive models
Machine-learned models that capture the microstructure impact on mechanical outcomes
Sep 26, 2021 12:00 AM — 12:00 AM San Diego, CA
See all events

Recent Publications

Quickly discover relevant content by filtering publications.
R. Jones, C. Safta, A. Frankel (2023). Deep learning and multi-level featurization of graph representations of microstructural data. Computational Mechanics.

Cite DOI

R. Wonnacott, D. Ching, J. Chilleri, C. afta, L. Rashkin, T. Reichardt (2023). Industrial PLC Network Modeling and Parameter Identification Using Sensitivity Analysis and Mean Field Variational Inference. Sensors.

Cite DOI

K. Kim, O. Diaz-Ibarra, H. Najm, J. Zador, C. Safta (2022). TChem: A performance portable parallel software toolkit for complex kinetic mechanisms. Computer Physics Communications.

Cite DOI

C. Safta, R. Ghanem, M. Grant, M. Sparapany, H. Najm (2022). Trajectory design via unsupervised probabilistic learning on optimal manifolds. Data-Centric Engineering.

Cite DOI

A. Hegde, E. Weiss, W. Windl, H. Najm, C. Safta (2022). Bayesian calibration of interatomic potentials for binary alloys. Computational Materials Science.

Cite DOI

P. Blonigan, J. Ray, C. Safta (2021). Forecasting Multi-Wave Epidemics Through Bayesian Inference. Archives of Computational Methods in Engineering.

Cite DOI

See all publications

Contact